About NVIDIA NeMo Relay

Release Notes

View as Markdown

This page summarizes the current documentation-visible release state. GitHub Releases remain the source of truth for complete release history and tag-specific notes.

Current Release

NVIDIA NeMo Relay 0.5 adds dynamic plugin SDKs and lifecycle controls, verified CLI installers and readiness diagnostics, stable token and cost field semantics, and stronger request and trace lineage behavior.

The most important compatibility notes are:

  • Plugin configuration now lives in plugins.toml. Move plugin settings out of config.toml; the --plugin-config CLI option has been removed.
  • Registered LLM request and tool execution intercepts now return canonical outcomes. Rebuild development native plugins and grpc-v1 workers against 0.5.
  • The experimental nemo-relay-wasm package, Cursor CLI support, and patch-based integrations have been removed.
  • Go and the raw C FFI remain experimental and source-first. Node.js 24 or newer remains the minimum supported Node.js version.

Use the child pages for release highlights and support notes. For the complete PR-by-PR changelog, release artifacts, and tag-specific history, use GitHub Releases.

Release Scope

This release includes:

  • Dynamic plugin discovery, trust policy and attestation gates, and lifecycle controls through the nemo-relay CLI.
  • A Rust native plugin SDK for trusted in-process rust_dynamic plugins, plus grpc-v1 worker support through Rust and Python SDKs.
  • Schema-aware dynamic-plugin configuration and editing through plugins.toml.
  • Verified Unix and Windows CLI installers, and expanded nemo-relay doctor readiness diagnostics for agent hosts and plugin configuration.
  • Canonical LLM request and tool execution intercept outcomes with ordered pending marks across the runtime and supported bindings.
  • Automatic Dynamo session-lineage headers on managed LLM requests and turn-scoped CLI trace correlation.
  • Stable normalized token and cost semantics for provider response codecs and ATOF, ATIF, OpenInference, and OpenTelemetry projections.

Feature Documentation

For the major 0.5 additions, start with: